Random splitting of datasets in image segmentation often leads to unrepresentative test sets, resulting in biased evaluations and poor model generalization. While stratified sampling has proven effective for addressing label distribution imbalance in classification tasks, extending these ideas to segmentation remains challenging due to the multi-label structure and class imbalance typically present in such data. Building on existing stratification concepts, we introduce Iterative Pixel Stratification (IPS), a straightforward, label-aware sampling method tailored for segmentation tasks. Additionally, we present Wasserstein-Driven Evolutionary Stratification (WDES), a novel genetic algorithm designed to minimize the Wasserstein distance, thereby optimizing the similarity of label distributions across dataset splits. We prove that WDES is globally optimal given enough generations. Using newly proposed statistical heterogeneity metrics, we evaluate both methods against random sampling and find that WDES consistently produces more representative splits. Applying WDES across diverse segmentation tasks, including street scenes, medical imaging, and satellite imagery, leads to lower performance variance and improved model evaluation. Our results also highlight the particular value of WDES in handling small, imbalanced, and low-diversity datasets, where conventional splitting strategies are most prone to bias.

Rheumatoid arthritis (RA) is a chronic inflammatory autoimmune disease affecting peripheral joints. Before clinical diagnosis, individuals may possess certain antibodies and experience discomfort but without specific signs of RA or inflamed joints. This stage is termed "preclinical RA," as these individuals are at risk of developing the disease. This early stage is difficult to define, necessitating the development of individual risk models. This study aims to estimate the time and risk of RA onset using various survival machine learning models. After identifying the best model, we stratify patients into risk categories and identify key risk factors. Data from 154 anonymized preclinical RA patients were collected and analyzed. Several survival analysis models were evaluated, including Survival Tree, Random Survival Forest, Extreme Gradient Boosting Survival, Linear Multi-Task Model, Neural Multi-Task Model, Support Vector Machines, and Cox Proportional Hazards. The Random Survival Forest model outperformed the others, achieving a mean C-index of 0.798. Using this model, patients were stratified into low-, medium-, and high-risk groups, facilitating personalized scheduling of clinical visits based on RA risk. To enhance model interpretability, SHapley Additive Explanations (SHAP) are employed to identify key risk factors. The baseline level of rheumatoid factor (RF) antibodies is the most significant predictor. Higher levels of anti-cyclic citrullinated peptide (anti-CCP) and RF antibodies at baseline are linked to earlier RA onset. This method provides valuable insights into key factors that might be overlooked in clinical practice and can improve patient management and quality of life for those at risk of developing RA.

Mixed-type time series (MTTS) is a bimodal data type that is common in many domains, such as healthcare, finance, environmental monitoring, and social media. It consists of regularly sampled continuous time series and irregularly sampled categorical event sequences. The integration of both modalities through multimodal fusion is a promising approach for processing MTTS. However, the question of how to effectively fuse both modalities remains open. In this paper, we present a comprehensive evaluation of several deep multimodal fusion approaches for MTTS forecasting. Our comparison includes different fusion types (early, intermediate, and late) and fusion methods (concatenation, weighted mean, weighted mean with correlation, gating, and feature sharing). We evaluate these fusion approaches on three distinct datasets, one of which was generated using a novel framework. This framework allows for the control of key data properties, such as the strength and direction of intermodal interactions, modality imbalance, and the degree of randomness in each modality, providing a more controlled environment for testing fusion approaches. Our findings show that the performance of different fusion approaches can be substantially influenced by the direction and strength of intermodal interactions. The study reveals that early and intermediate fusion approaches excel at capturing fine-grained and coarse-grained cross-modal features, respectively. These findings underscore the crucial role of intermodal interactions in determining the most effective fusion strategy for MTTS forecasting.

Localization based on channel impulse responses (CIRs) of radio frequency (RF) signals yields centimeter-accurate positions under optimal line-of-sight (LOS) propagation conditions. However, in real indoor environments, e.g., in car manufacturing, non-line-of-sight (NLOS) situations dominate. Here, multipath propagation affects the time-of-arrival (ToA) estimation and downstream multilateration and localization accuracy. The detection and subsequent mitigation of NLOS per transceiver line compensates for these effects. To detect NLOS, the state-of-the-art employs either supervised or unsupervised learning methods that require the acquisition of expensive reference data or do not generalize to changes or unknown environments. This is due to, among other things, the fact that they cannot exploit spatial and temporal information from CIR signal streams.Thus, we propose a generative deep state space model (SSM) for NLOS detection on CIRs that exploits time and space. Our ultra-wideband (UWB) experiments show that our dynamical variational autoencoder (DVAE) detects NLOS signals from sequences of CIRs more accurately than the state-of-the-art and is robust to unknown environments.

Certified defenses against adversarial attacks offer formal guarantees on the robustness of a model, making them more reliable than empirical methods such as adversarial training, whose effectiveness is often later reduced by unseen attacks. Still, the limited certified robustness that is currently achievable has been a bottleneck for their practical adoption. Gowal et al. and Wang et al. have shown that generating additional training data using state-of-the-art diffusion models can considerably improve the robustness of adversarial training. In this work, we demonstrate that a similar approach can substantially improve deterministic certified defenses but also reveal notable differences in the scaling behavior between certified and empirical methods. In addition, we provide a list of recommendations to scale the robustness of certified training approaches. Our approach achieves state-of-the-art deterministic robustness certificates on CIFAR-10 for the $\ell_2$ ($\epsilon = 36/255$) and $\ell_{\infty}$ ($\epsilon = 8/255$) threat models, outperforming the previous results by $+3.95$ and $+1.39$ percentage points, respectively. Furthermore, we report similar improvements for CIFAR-100.

The widespread success of convolutional neural networks may largely be attributed to their intrinsic property of translation equivariance. However, convolutions are not equivariant to variations in scale and fail to generalize to objects of different sizes. Despite recent advances in this field, it remains unclear how well current methods generalize to unobserved scales on real-world data and to what extent scale equivariance plays a role. To address this, we propose the novel Scaled and Translated Image Recognition (STIR) benchmark based on four different domains. Additionally, we introduce a new family of models that applies many re-scaled kernels with shared weights in parallel and then selects the most appropriate one. Our experimental results on STIR show that both the existing and proposed approaches can improve generalization across scales compared to standard convolutions. We also demonstrate that our family of models is able to generalize well towards larger scales and improve scale equivariance. Moreover, due to their unique design we can validate that kernel selection is consistent with input scale. Even so, none of the evaluated models maintain their performance for large differences in scale, demonstrating that a general understanding of how scale equivariance can improve generalization and robustness is still lacking.

Popular metrics for clustering comparison, like the Adjusted Rand Index and the Adjusted Mutual Information, are type II biased. The Standardized Mutual Information removes this bias but suffers from counterintuitive non-monotonicity and poor computational efficiency. We introduce the $p$-value adjusted Rand Index ($\operatorname{PMI}_2$), the first cluster comparison method that is type II unbiased and provably monotonous. The $\operatorname{PMI}_2$ has fast approximations that outperform the Standardized Mutual information. We demonstrate its unbiased clustering selection, approximation quality, and runtime efficiency on synthetic benchmarks. In experiments on image and social network datasets, we show how the $\operatorname{PMI}_2$ can help practitioners choose better clustering and community detection algorithms.

Ultra-wideband (UWB) systems based on Channel State Information (CSI) estimate the position of mobile nodes within an environment by using Channel Impulse Responses (CIRs) of multiple stationary nodes. These contain spatial information caused by environment interactions such as reflections and scattering. To estimate positions from CSI of stationary nodes, we must transmit them to a centralized node. This introduces considerable communication overhead.We present a large-scale study to determine whether CSI can be compressed into a small set of underlying latent variables that describe the most valuable information. We evaluate multiple neural network architectures containing encoding (compressing) and decoding (reconstructing) components and compare them to the state-of-the-art compression techniques Discrete Cosine Transform (DCT) and Discrete Wavelet Transform (DWT). We show that fully connected autoencoders achieve the lowest error, outperforming both DCT and DWT. Further experiments prove that the reconstructed CSI can be used for positioning with only mild performance deterioration at a compression of >97% and even when trained on a different environment.